Fix sp non-padding-free

[tastelikefeet]

PR type

• Bug Fix
• New Feature
• Document Updates
• More Models or Datasets Support

PR information

Write the detail information belongs to this PR.

Experiment results

Paste your experiment result here(if needed).

[gemini-code-assist]

Summary of Changes

Hello @tastelikefeet, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request significantly expands the system's capabilities for reranker and sequence classification tasks, particularly focusing on improved integration with the vLLM inference engine. It introduces a new control for reranker activation, adds support for a popular family of reranker models, and provides practical deployment examples. Additionally, it includes a critical fix for sequence parallelism, enhancing overall stability and functionality.

Highlights

• Reranker Activation Control: Introduced a new argument, reranker_use_activation, allowing users to control whether a sigmoid activation function is applied after reranker scores. This provides more flexibility in handling reranker outputs.
• BGE Reranker Model Support: Added comprehensive support for BAAI's BGE reranker models (e.g., bge-reranker-base, bge-reranker-v2-m3, bge-reranker-large), including their model types and templates for seamless integration.
• Enhanced VLLM Integration for Reranker and Sequence Classification: The vLLM inference engine has been significantly updated to support seq_cls, reranker, and generative_reranker tasks, including proper handling of pooling parameters and response creation for these specific task types.
• New Deployment Examples: Provided new client and server scripts for deploying reranker and sequence classification models using vLLM, simplifying the process for users to set up and test these models.
• Documentation Updates: Updated both English and Chinese documentation to reflect the new reranker_use_activation parameter and to list the newly supported BGE reranker models.
• Sequence Parallelism Fix: Resolved a bug in ulysses.py related to the handling of position_ids when sequence parallelism is enabled, ensuring correct behavior in such scenarios.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request primarily introduces support for reranker models, including BGE and Qwen3 generative rerankers, and enhances sequence classification capabilities with the vLLM backend. It also contains a fix for sequence parallelism in non-padding-free mode. The changes are comprehensive, including necessary documentation updates and new examples. My review includes a couple of suggestions to improve code modularity and clarity, mainly by encapsulating model-specific logic within template classes.